Bite-Mark Comparison: Two Premises Under Test

They do not vary independently. Teeth crowd and lean together, so their positions are correlated, and multiplying correlated quantities as if they were independent inflates the rarity. Mary Bush, Peter Bush and David Sheets tested it directly in 2011, in a paper titled to answer Rawson: "Statistical Evidence for the Similarity of the Human Dentition." They measured two sets of scanned dental models, 172 in one and 344 in the other, and counted how often the six lower front teeth, the ones that usually leave a mark, came out alike within measurement tolerance. They found seven matches in the smaller set and sixteen in the larger. Their conclusion was that in an open population, meaning the whole field of possible biters rather than a short list of suspects, statements of dental uniqueness for bite-mark purposes are unsupportable.

The official reviews reached the same place. The 2009 National Academy of Sciences report listed both premises as scientifically unestablished, and the 2023 NIST scientific-foundation review by Kelly Sauerwein and colleagues, which considered more than 400 sources, found the field’s key premises unsupported by data. PCAST, reviewing the same field in 2016, went further than it did for any other method it examined: it judged the prospects of ever validating bite-mark analysis to be low, and advised against devoting significant resources to the attempt. The measured counter-view is worth stating: Franco and colleagues (2017), examining 445 casts including twins and orthodontic cases, likewise found no evidence for uniqueness, but argued bite-mark work "should not be disregarded" where a closed group of suspects has distinctive dental features. And even sharp critics accept the teeth may well be unique. Uniqueness, though, is only the first premise, and on its own it identifies no one.

The result repeats. Bush and colleagues (2011) impressed 89 marks from one dentition into cadaver skin and none corresponded to the biting teeth within measurement error. The warning is not new. Duane DeVore reported in 1971 that skin marked with an ink grid stretched by up to 60 percent as the body changed position, and that a marked area shrank by about 45 percent once excised and preserved. He concluded that a superimposition comparison is invalid unless the exact body position at the moment of the bite is known and reproduced.

Two honest qualifications belong here. DeVore also allowed that a distinctive or abnormal dentition might still be distinguishable, so the failure is not absolute. And the Bush studies used unembalmed cadaver skin, which does not bruise or swell; a living victim’s response is more likely to add distortion than to remove it. The conclusion the experiments support is narrow and firm: skin does not reliably carry the detail of the dentition, and a measurement taken from a bitten area is partly a measurement of the body and its position, not only of the teeth.

Agreement on the comparison itself is limited, though not zero, and the fairer studies deserve their due. Kristopher Arheart and Iain Pretty (2001) had 32 diplomates work a small set of clear, higher-value cases and reached a combined accuracy, measured as area under the ROC curve, of 0.86. The authors called that "less than optimal," and found accuracy was unrelated to years of experience. Iain Pretty and David Sweet (2001) measured inter-examiner reliability on the standard overlay technique and found it only moderate, with board certification giving no advantage. And in a controlled animal-model study, Avon and colleagues (2010) found that even diplomates, as a group and individually, made critical errors, attributing a bite to a suspect who had not made it.

The honest summary is mixed rather than uniformly damning. On clear marks and pooled judgements the field can perform moderately. What it cannot do is agree reliably on the first question, whether a given injury is even a bite, and the examiner’s years of experience do not predict who gets the answer right.

The first empirical test came from Nikola Osborne, Sally Woods, Jules Kieser and Rachel Zajac (2014). Dental and non-dental students judged bite-mark pairs, some clear and some deliberately ambiguous, while the emotional context was manipulated. Context changed decisions only on the ambiguous pairs, and in an unexpected direction, producing fewer same-source calls rather than more. The participants were students and the marks were simulated, so the size and direction of the real-world effect remain open. But the location of the effect is the familiar one: ambiguous evidence is where context does its work, and a distorted bite mark is ambiguous evidence.

The safeguards are procedural, not attitudinal. Examine the mark and record a conclusion before seeing the suspect’s cast; release case information in sequence rather than all at once; use a second examiner who works without the case theory or the first opinion. What counts in court is whether those steps were taken, not an assurance that the examiner set the context aside. An opinion formed after absorbing the case is not independent of it.

The reliable question turned out to be a different one. A bite deposits the biter’s saliva, and DNA reads it. David Sweet and colleagues (1997) recovered typable DNA from saliva left on skin; Sweet and Gary Shutler (1999) did it on a body recovered from a river, and used the profile to exclude 33 of 34 suspects, where the physical comparison could offer only "probable." By 2013 the Bush group was noting that a bite site might be better used as a source of DNA than as a pattern to compare.

The balanced view still leaves the mark a role. It can show where to swab. It can sometimes exclude a dentition, which may clear a suspect. Pretty and Sweet point out that bite marks have helped exonerate the innocent and protect children, and Franco’s work suggests a distinctive dentition in a closed group of suspects carries more information than an ordinary one. But when an identification is possible at all, the load-bearing evidence now comes from the saliva, not the shape.

Steven Chaney was convicted on a "one in a million" bite-mark claim that the expert, Jim Hales, later recanted in a sworn affidavit. Robert Lee Stinson lost 23 years on an identification given with "no margin for error," confirmed as a second opinion by the same Raymond Rawson who had testified against Krone, until DNA identified Moses Price. William Richards was convicted on a bite-mark opinion the odontologist, Norman Sperber, later repudiated; California then rewrote its law so that repudiated or scientifically undermined expert testimony counts as false evidence, and the conviction was vacated in 2016.

Roy Brown spent fifteen years inside on a bite-mark opinion from a local dentist with no board certification and no formal training in the analysis — after a qualified odontologist had already examined the same marks and concluded Brown could be excluded. The prosecution set that exclusion aside and commissioned the non-expert opinion that identified him. Brown obtained the buried report under freedom-of-information law, named the real biter himself, and DNA confirmed it in 2007. Pretty and Bowers, whose account this draws on, use the case to make the uncomfortable point that the correct science was available the whole time and the process discarded it.

Three lessons run through the set. The correct science can be available and still discarded, as Brown’s exclusion was. Admissibility never established validity: a Mississippi court had affirmed Brooks’s conviction and, in doing so, declared bite-mark evidence admissible in the state. And the counts kept climbing. Michael Bowers (2019) recorded at least 31 exonerations, dismissals and releases tied to bite-mark evidence; by 2023 the figure cited within the field was 38. Even Michael West, the discipline’s most notorious practitioner, came to say in a sworn deposition that bite-mark analysis should be thrown out of court. The working lesson is plain: know these cases better than counsel does, say only what the mark supports, and where an identification matters, swab for DNA.